A year after Agility Robotics’ humanoid robot Digit began working in a warehouse, lifting boxes of Spanx, the conversation around humanoid robots has shifted from novelty to necessity—particularly the need for clear safety standards. Digit, which can lift up to 16 kilograms and stands 1.8 meters tall, performs heavy lifting tasks normally handled by humans. But despite its remarkable capabilities, Digit is still prone to falling, as it did during a trade show in March when it unexpectedly collapsed, dropping the box it was carrying and face-planting onto the concrete floor.
As reported by MIT Technology Review, this kind of malfunction underscores why experts are urgently calling for safety protocols tailored specifically for humanoid robots. Unlike traditional robots that operate on fixed bases or wheels, humanoids are dynamically stable—they need power just to stay upright. This makes emergency shut-off procedures a complex safety concern. “If a robot were to hit your throat, even with a fraction of the force needed to lift a 50-pound tote, it could seriously injure a person,” says Pras Velagapudi, CTO of Agility Robotics.
The IEEE Humanoid Study Group, which shared its initial findings with MIT Technology Review, emphasizes that humanoids require a new set of safety standards. Existing frameworks don’t account for the physical and psychological challenges posed by robots designed to share space and interact directly with humans. These robots will eventually be working in homes, hospitals, and public places—not just behind laser barriers in warehouses—and the risk of injury or misunderstanding grows significantly in those settings.
Agility Robotics is already responding with updates to Digit that allow it to decelerate gradually and lower itself into a safe posture rather than powering down abruptly. Other companies, like Boston Dynamics, are also involved in shaping international standards through groups like the ISO. Federico Vicentini of Boston Dynamics points out that the industry needs to agree on what the safety goals are, while allowing designers the freedom to innovate on how to meet them.
One of the challenges in creating universal standards is defining what even qualifies as a humanoid robot. Does it need arms and legs? A face? The IEEE group suggests moving away from the term entirely and instead classifying robots based on behavior, capabilities, and use cases. This would ensure that safety standards apply broadly to everything from bipedal bots to quadrupeds like Spot.
Aside from the physical concerns, humanoid robots also pose psychosocial risks. Because these machines resemble people, users tend to anthropomorphize them, expecting levels of interaction they aren’t equipped to deliver. This misalignment between appearance and capability can lead to frustration or even unsafe behavior. Prather from the IEEE notes, “Sometimes you let your guard down on safety, or your expectations of what that robot can do versus reality go higher.”
Emotional safety is a key focus of the IEEE’s upcoming report. Surveys conducted by Greta Hilburn at the U.S. Defense Acquisition University found that people expect humanoid robots to communicate with facial expressions, gestures, and emotional sensitivity—features that currently don’t exist. These unrealistic expectations could cause real harm, especially in settings involving children, elderly individuals, or people with disabilities.
As robots move beyond warehouses into daily human environments, it becomes critical that they’re equipped with standardized cues to communicate their actions—like brake lights on a car. Digit already uses lights to indicate its direction, but audio cues aren’t practical in loud environments, and multiple robots could easily cause confusion. Ensuring that all people, including those with hearing or vision impairments, can safely interact with these machines is another layer of complexity.
The IEEE group’s recommendations include enabling human overrides, aligning robot appearance with function, and creating inclusive communication methods. Prather emphasizes that these standards aren’t about limiting progress—they’re about ensuring safety and building trust. “We’re just saying you’ve got to hit this minimum bar—and we all agree below that is bad,” he says.
As MIT Technology Review notes, the industry is still in early stages. But before humanoid robots become a regular part of our daily lives, manufacturers, regulators, and designers must agree on what safe, responsible robotics looks like. Otherwise, the very tools designed to help us could become liabilities instead.
